A series of experiments in a thin layer geometry performed at the HYKA test site of the KIT. The experiments on different combustion regimes for lean and stoichiometric H2/air mixtures were performed in a rectangular chamber with dimensions of 20 x 90 x h cm3 , where h is the thickness of the layer (h = 1, 2, 4, 6, 8, 10 mm). Three different layer geometries: (1) a smooth channel without obstructions; (2) the channel with a metal grid filled 25% of length and (3) a metal grid filled 100% of length. Detail measurement of H2/air combustion behavior including flame acceleration (FA) and DDT in closed rectangular channel have been done. Five categories of flame propagation regimes were classified. Special attention was paid to the analysis of critical condition for different regimes of flame propagation as function of the layer thickness and roughness of the channel. It was found that thinner layer suppresses the detonation onset and even with a roughness, the flame is available to accelerate to speed of sound. The detonation may occur only in a channel thicker than 6 mm.

A study with PIV-measurements for gas explosion in hydrogen-air mixtures is presented in this paper. The present work is part of an ongoing research project. The experiments are performed with hydrogen-air mixture at atmospheric pressure and room temperature. The experimental rig is a square channel with 4.5 X 2.0 cm2 cross section, 30 cm long with a single cylindrical obstacle of blockage ratio 1/3. The equipment used for the PIV-measurements was a Firefly diode laser from Oxford lasers, Photron SA-Z high speed camera and a particle seeder producing 1 µm droplets of water. The gas concentrations used in the experiments was between 14 and 17% hydrogen in air. The resulting explosion can be characterized as slow. Explosions in the gas mixtures at the highest hydrogen concentration produced measured flow velocities of up to 17 m/s as the flame passed the obstacle. Similar velocities was also measured one channel height behind the obstacle. The flow vortices produced behind the obstacle seemed to give separation between the liquid droplets and gas flow. The experimental results can be used for reference in validation of CFD-codes.

The numerical simulations on the high-pressure hydrogen jet are performed by using the unsteady threedimensional compressible Navier-Stokes equations with multi-species conservation equations. The present numerical results show that the highly expanded hydrogen free jet observes and the distance between the Mach disc and the nozzle exit agrees well with the empirical equation. The time-averaged H2 concentration of the numerical simulations agrees well with the experimental data and the empirical equation. The numerical simulation of ignition in a hydrogen jet is performed to show the flame behaviour from the calculated OH isosurface. We predicted the ignition and no-ignition region from the present numerical results about the forced ignition in the high-pressurized hydrogen jet.

The formation of hydrogen jets from pressurized sources and its ignition when hitting hot devices has been studied by many projects. The transient jets evolve with high turbulence depending on the configuration of the nozzle and especially the pressure in the hydrogen reservoir. In addition the length of the jets and the flames generated by ignition at a hot surface varies. Parameters to be varied were initial pressure of the source (2.5, 10, 20 and 40 MPa), distance between the nozzle and the hot surface (3, 5 and 7 m) and temperature of the hot surface (between 400 and 1000 K). The interaction of the hydrogen jets is visualized by high-speed cinematography techniques which allow analysing the jet characteristics. By combination of various methods of image processing, the visibility of the phenomena on the videos taken at 15 000 fps was improved. In addition, high-speed NIR spectroscopy was used to obtain temperature profiles of the expanding deflagrations. The jets ignite already above 450 K for conditions mainly from the tubular source at 40 MPa. In addition, the propagation of the flame front depends on all three varied parameters: temperature of the hot surface, pressure in the reservoir and distance between nozzle and hot surface. In most cases also upstream propagation occurs. A high turbulence seems to lead to the strong deflagrations. At high temperatures of the ignition sources, the interaction leads to fast deflagration and speeds up- and downstream of the jet. The deflagration velocity is close to velocity of sound and emission of pressure waves occurs.

In the current study, a Large Eddy Simulation strategy is applied to model the dispersion of compressible turbulent hydrogen jets issuing from realistic pipe geometries. The work is novel, as it explores the effect of jet densities and Reynolds numbers on vertical buoyant jets, as they emerge from the outer wall of a pipe, through a round orifice, perpendicular to the mean flow within the pipe. An efficient Godunov solver is used, and coupled with Adaptive Mesh Refinement to provide high resolution solutions only in areas of interest. The numerical results are validated against physical experiments of air and helium, which allows a degree of confidence in analysing the data obtained for hydrogen releases. The results show that the jets investigated are always asymmetric. Thus, significant discrepancies exist when applying conventional round jet assumptions to determine statistical properties associated with gas leaks from pipelines.

The growing number of hydrogen fillings stations and cars increases the need for accurate models to determine risk. The effect on hydrogen flame length was measured by varying the diameter of the spouting nozzle downstream from the chocked nozzle upstream. The results was compared with an existing model for flame length estimations. The experimental rig was setup with sensors that measured accurately temperature, mass flow, heat radiation and the pressure range from 0.1 to 11 MPa. The flame length was determined with an in-house developed image-processing tool, which analyzed a high-speed film of the each experiment. Results show that the nozzle geometry can cause a deviation as high as 50% compared to estimated flame lengths by the model if wrong assumptions are made. Discharge coefficients for different nozzles has been calculated and presented.

Thermal radiation from an under-expanded (900 bar) hydrogen jet fire has been numerically investigated. The simulation results have been compared with the flame length and radiative heat flux measured for the horizontal jet fire experiment conducted at INERIS. The release blowdown characteristics have been modelled using the volumetric source as an expanded implementation of the notional nozzle concept. The CFD study employs the realizable κ-ε model for turbulence and the Eddy Dissipation Concept for combustion. Radiation has been taken into account through the Discrete Ordinates (DO) model. The results demonstrated good agreement with the experimental flame length. Performance of the model shall be improved to reproduce the radiative properties dynamics during the first stage of the release (time < 10 s), whereas, during the remaining blowdown time, the simulated radiative heat flux at five sensors followed the trend observed in the experiment.

A hydrogen leak from a facility, which uses highly compressed hydrogen gas (714 bar, 800 K) during operation was studied. The investigated scenario involves supersonic hydrogen release from a 10 cm2 leak of the pressurized reservoir, turbulent hydrogen dispersion in the facility room, followed by an accidental ignition and burn-out of the resulting H2-air cloud. The objective is to investigate the maximum possible flame velocity and overpressure in the facility room in case of a worst-case ignition. The pressure loads are needed for the structural analysis of the building wall response. The first two phases, namely unsteady supersonic release and subsequent turbulent hydrogen dispersion are simulated with GASFLOW-MPI. This is a well validated parallel, all-speed CFD code which solves the compressible Navier-Stokes equations and can model a broad range of flow Mach numbers. Details of the shock structures are resolved for the under-expanded supersonic jet and the sonic-subsonic transition in the release. The turbulent dispersion phase is simulated by LES. The evolution of the highly transient burnable H2-air mixture in the room in terms of burnable mass, volume, and average H2-concentration is evaluated with special sub-routines. For five different points in time the maximum turbulent flame speed and resulting overpressures are computed, using four published turbulent burning velocity correlations. The largest turbulent flame speed and overpressure is predicted for an early ignition event resulting in 35 -71 m/s, and 0.13 – 0.27 bar, respectively.